Module 2 • Lesson 13• 55 mins
Navigate Implementation Strategy Across Buy, RAG, Skills, and Fine-Tune
Navigate the Customization Spectrum from Buy (APIs) to Customize (RAG vs Skills) and Fine-tune; execute comprehensive trade-off audits and define escalation triggers.
4 tiers on the spectrum: Buy, RAG, Skills, Fine-tune
Disambiguate RAG (Dynamic document retrieval) vs Skills (Codified expert SOPs)
The Start-with-Buy principle and explicit technical escalation triggers
Navigate Implementation Strategy Across Buy, RAG, Skills, and Fine-Tune
Following the task risk diagnosis from Lesson 8, the subsequent architectural decision is: How should we technically implement this AI capability? In AI product management, this is not a simplistic binary "Build vs. Buy" decision, but a nuanced Customization Spectrum.
Running Example: Tech recruiting platform FinTrack Talent developing an automated feature: "AI parses a candidate's CV and Job Description (JD) to generate 3 targeted interview questions probing specific competency gaps."
TIẾP CẬN CŨ / LỖI THỜI
Binary Mindset: Build vs. Buy
Overlooks high-leverage contextual customization layers
CHUẨN AI PM / HIỆN ĐẠI
AI Customization Spectrum: 4 Tiers
Buy (API) → Customize (RAG / Skills) → Fine-tune → Build from Scratch
AI UNIT ECONOMICSToken Costs, Latency & Model Cascading
Calculate AI Unit Economics: Token Costs, Latency, and True ROI
Monthly Query Volume:100,000 queries
Model Architecture:
Human Escalation Rate:2%
LLM TOKEN BILL
$1,500
HUMAN AUDIT COST
$1,600
UNIT COST / QUERY
$0.0313
GROSS MARGIN
61%